I feel like aspell is one of those programs that could benefit from some kind of effort to crowdfund necessary open source packages. It needs someone to show it some love, even if that someone doesn’t have a passion about working on spell checkers but is getting paid good money to look after it part-time.
The comparison on aspell.net seems to be only for English – is aspell better than the alternatives for any other languages?
There’s also voikko, which started out only for Finnish (what is it about Uralic languages and spelling engines …), but now supports pretty much anything through the use of HFST, using weighted finite state transducers and Levenshtein automatons. HFST can be used as pretty much a drop-in replacement for Xerox Finite State Tools, which is/was the default toolkit for linguists writing computational morphologies in academia (probably has seen some industry use too). If we’re playing the long game, it seems like a better strategy to create an HFST-readable dictionary + error model for English that voikko can use, and try to get it to aspell’s level, which should open for better spellers for all languages. Voikko and HFST are under active development, and there’s lots of people creating language data using HFST (e.g. the Apertium project and University of Tromsø and of course University of Helsinki).